Search papers, labs, and topics across Lattice.
The paper introduces Multi-scale Weak-form Physics-Informed Neural Networks (MUSA-PINN) to improve the convergence and accuracy of PINNs for solving fluid flow problems in complex geometries like Triply Periodic Minimal Surfaces (TPMS). MUSA-PINN reformulates PDE constraints as integral conservation laws over hierarchical spherical control volumes at three scales, enforcing continuity and momentum conservation via flux-balance residuals. Experiments demonstrate that MUSA-PINN significantly outperforms state-of-the-art PINN baselines, achieving up to 93% reduction in relative errors and improved mass conservation in TPMS geometries.
Standard PINNs stumble in complex geometries, but MUSA-PINN leaps ahead by reformulating PDE constraints as multi-scale integral conservation laws, slashing errors by up to 93% in fluid flow simulations.
While Physics-Informed Neural Networks (PINNs) offer a mesh-free approach to solving PDEs, standard point-wise residual minimization suffers from convergence pathologies in topologically complex domains like Triply Periodic Minimal Surfaces (TPMS). The locality bias of point-wise constraints fails to propagate global information through tortuous channels, causing unstable gradients and conservation violations. To address this, we propose the Multi-scale Weak-form PINN (MUSA-PINN), which reformulates PDE constraints as integral conservation laws over hierarchical spherical control volumes. We enforce continuity and momentum conservation via flux-balance residuals on control surfaces. Our method utilizes a three-scale subdomain strategy-comprising large volumes for long-range coupling, skeleton-aware meso-scale volumes aligned with transport pathways, and small volumes for local refinement-alongside a two-stage training schedule prioritizing continuity. Experiments on steady incompressible flow in TPMS geometries show MUSA-PINN outperforms state-of-the-art baselines, reducing relative errors by up to 93% and preserving mass conservation.
